one last thing - you could rename SjaakII directory that holds VS2012 solution and project to something like build_vs2012.
Best would be of course to use cmake to generate these but unfortunately I can't help with that.
as for the changes in compilerdef.h: either use my changes or include
compilerdef.h as first file (this just to get rid of one warning).
Could anyone make me a Windows compile...?
Moderators: hgm, Rebel, chrisw
-
- Posts: 2555
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: Could anyone make me a Windows compile...?
Same here, forgot to mention this.mar wrote: benchmark is 16160856 nodes.
I've renamed the "SjaakII" subdirectory to "MSVS2012". The files are text files, so I should be able to update them "by hand" to keep them current. They have some hard-coded paths in them though, so I wonder how portable they are? It's possible to generate them from CMake (according to the documentation anyway) but I think you need to be on Windows with MSVS installed (CMake is meant to configure the project for compilation on the current system, not to produce distributable build files).
Many thanks for your help! Quick question: is the binary 32 or 64 bit?
-
- Posts: 2555
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: Could anyone make me a Windows compile...?
You're welcome. There should be no absolute paths, I fixed that already in the project (maybe I'm missing something?).
You mean the binary I ran test bechmark with?
That was the 64-bit binary, but I've done a quick fixed depth search for both versions and node counts match, so I'm pretty sure both binaries are ok.
There's still the hash problem with 32-bit binary so test benchmark still crashes, but I guess changing hash size on the fly should do.
You mean the binary I ran test bechmark with?
That was the 64-bit binary, but I've done a quick fixed depth search for both versions and node counts match, so I'm pretty sure both binaries are ok.
There's still the hash problem with 32-bit binary so test benchmark still crashes, but I guess changing hash size on the fly should do.
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: Could anyone make me a Windows compile...?
No, you're right. I mis-read the diff. The lines that have absolute paths are all deleted.mar wrote:You're welcome. There should be no absolute paths, I fixed that already in the project (maybe I'm missing something?).
I actually meant the binary in the archive. It's identified as "PE32+ executable for MS Windows (console) Mono/.Net assembly". No idea "PE32+" means exactly but my first inclination is to read that as "32 bit"; either way, if you have a 32 and a 64 bit version that I can include in the archive that would be great.You mean the binary I ran test bechmark with?
That was the 64-bit binary, but I've done a quick fixed depth search for both versions and node counts match, so I'm pretty sure both binaries are ok.
There's still the hash problem with 32-bit binary so test benchmark still crashes, but I guess changing hash size on the fly should do.
I forgot about the hashtable size. I suppose I should just reduce the default size; if I do it selectively for 32 and 64 bit systems it's probably going to change the bench result between the two versions.
-
- Posts: 2555
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: Could anyone make me a Windows compile...?
There should be two binaries. I'll repost the links:
http://www.crabaware.com/sjaak/sjaakII_intel_32bit.zip
http://www.crabaware.com/sjaak/sjaakII_intel_64bit.zip
the 64-bit one is a bit smaller.
http://www.crabaware.com/sjaak/sjaakII_intel_32bit.zip
http://www.crabaware.com/sjaak/sjaakII_intel_64bit.zip
the 64-bit one is a bit smaller.
-
- Posts: 2555
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: Could anyone make me a Windows compile...?
I guess it's not a big deal. However if you decide to reduce the size I would opt for 1/4 of the default valueEvert wrote:I forgot about the hashtable size. I suppose I should just reduce the default size; if I do it selectively for 32 and 64 bit systems it's probably going to change the bench result between the two versions.
EDIT: if you decide to change the default hash size, let me know and I'll build beta6_r181
-
- Posts: 7218
- Joined: Mon May 27, 2013 10:31 am
Re: Could anyone make me a Windows compile...?
100 ELO means two or three year work in the old days. I you can add one ELO point per week you are one of the best.
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: Could anyone make me a Windows compile...?
I'm pretty sure it's always been possible to lose hundreds of elo points in less than five minutes of work by introducing a bug.
-
- Posts: 27796
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Could anyone make me a Windows compile...?
OK, I just made a silent update of the mini-Shogi package, putting in the latest Sjaak-II compiles from Martin.
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: Could anyone make me a Windows compile...?
I'm getting ready to push out the next release (it seems like the most obvious bugs have been fixed) but a question came up (from the Debian side) that needs to be clarified.
The code for the snprintf() replacement
did you write it yourself? Is it ok to stick a GPL license on it?
I assume yes in both cases, but I need to know for sure.
The code for the snprintf() replacement
Code: Select all
int snprintf(char *buf, size_t size, const char *fmt, ...)
{
int res;
va_list ap;
va_start(ap, fmt);
res = vsnprintf(buf, size, fmt, ap);
va_end(ap);
return res;
}
I assume yes in both cases, but I need to know for sure.